Development of a system-wide pharmacy operational weighted workload model at a large academic health system

Abstract

Purpose

The purpose of this study is to develop a standard operational and distributional weighted workload model that is applicable across an integrated, diverse healthcare system. This model aims to not only demonstrate the operational intensity of pharmacy practice but also to inform opportunities to decrease waste, increase efficiency, facilitate growth, and demonstrate value across operational and distributional pharmacy services.

Summary

Time studies were conducted at 8 hospitals within the UNC Health system to objectively measure time spent within each operational process in order to create a system-wide weighted workload model. Time study results informed the development of a system-wide weighted workload model. Data from December 29, 2019, through December 26, 2020, was then applied to this weighted workload model. With this model, acute care hospital and infusion center operational areas were compared in thousands of combinations within single operational areas and across any and all operational areas by dispense code, weighted work, and ratio of weighted work to total sum of dispenses at each site.

Conclusion

The model successfully achieved the objective to develop a standard operational weighted workload model that is applicable across the integrated, diverse care system. This model provides a foundation for UNC Health to further productivity measurement and fills a gap in the literature by offering a novel method of developing a system-level operational workload model that can be used to evaluate and compare operational workloads across health-system sites.

Key Points

• Current operational weighted workload and productivity models are limited to a single site and may not be applicable across a large and diverse health system.

• Creating a weighted workload model will comprehensively capture work tied to a dispense code across varying practice settings.

• A weighted workload model will serve to better articulate the distinct differences between operational areas and can help visualize changes that occur operationally over time.

Operational workload metrics and benchmarking are common in many industries and are being increasingly used in healthcare to measure departmental success and improve departmental performance. Operational workload metrics specific to pharmacy have been discussed in the literature; however, a notable limitation of these studies is that they have only been conducted at a single site.^1-3 A study done at the University of Wisconsin outlined productivity indicators that should be considered when developing a pharmacy operational model.^1,2 These indicators included worked hours per unit of service, drug cost per admission, labor cost per admission, total cost per admission, doses dispensed per admission, labor expense per 1,000 doses billed, pharmacist worked hours per order, technician worked hours per dose, inventory turns per year, clinical interventions per pharmacist shift worked, and pharmacy cost as a percentage of total hospital costs.^1,2 Another study was done to determine relative work during a 1-hour timeframe at Cleveland Clinic’s oncology pharmacy, where relative value units as weights were created and assigned to activities within a single satellite location.³

Applying single-site workload models throughout a health system may not accurately represent pharmacy practice at the various sites, given the diversity of hospitals within the health system (eg, variation in hospital size, automation practices, patient acuity, and clinical practice models).⁴ For example, University of North Carolina (UNC) Health is a diverse health system consisting of 11 hospitals that span the state of North Carolina. The hospitals and affiliated infusion centers are organized regionally, with 3 hospitals being in the system’s Statewide West Region, 4 hospitals in the Statewide East Region, 2 hospitals in the Triangle East Region, and 2 hospitals in the Triangle West Region. Similar to other health systems nationally and internationally, these hospitals differ in bed capacity and patient acuity. Specifically, the health system includes a critical access hospital, multiple community hospitals, and a large academic medical center, which is home to a level 1 trauma center, a National Cancer Institute–designated Comprehensive Cancer Center, a designated Pediatric Center of Excellence, a regional hemophilia center, and a regional burn center. While all hospitals within the system are not-for-profit facilities, 5 hospitals are owned entities and 6 are operated under a management services agreement. Further, all hospitals utilize a variety of different technologies within their central and decentralized pharmacy spaces.

As health systems continue to grow and diversify across the nation, more quantifiable information on best practices is needed for measuring workload productivity at the health-system level that is inclusive and accounts for the diversity of its hospitals. A critical first step in developing a health-system workload model is to identify workload processes and tasks that occur across sites within the health system. Zeeman and colleagues⁴ accomplished this with a focus on operational workload using a modified Delphi methodology. Delphi methodology is a research methodology that aims to develop consensus on a complex topic for which there is limited or conflicting evidence through a facilitated, iterative feedback and discussion process using a panel of experts in the field.^5-8 The Delphi expert panel identified and reached consensus on operational workload processes and corresponding tasks occurring at various hospitals within the diverse health system.⁴

Considering the distinctiveness of the pharmacies within UNC Health, a need to create a weighted workload model at the health-system level to begin to evaluate the metrics that factor into the work output per labor input productivity calculation was identified. The objective of this study described here was to develop a standard pharmacy operational weighted workload model that is applicable across an integrated, diverse healthcare system. The authors hypothesized that the development of this model would not only demonstrate the dynamic operational intensity of pharmacy practice but contribute to the creation of a tool that is able to decrease waste, increase efficiency, facilitate growth, and demonstrate value across operational and distributional pharmacy services. Further, the process of developing this weighted workload model provides a generalizable approach and model for pharmacy leaders across the world as a focus on decreasing costs and reducing waste continues to be important.

Methods

This study included 8 of the current 11 UNC Health hospitals; 2 hospitals were excluded as they did not yet use the system standard electronic health record (EHR) leveraged by the other 9 hospitals, and 1 hospital was excluded as it was added to the health system after study initiation. A modified Delphi method using an expert panel of operational leaders across the health system was conducted to identify and reach consensus on processes that occur in operational areas.⁴ Further detail on the Delphi method is described elsewhere.⁴ This study was submitted to the UNC institutional review board and designated as non–human subjects research.

Upon consensus of the operational workload process categories and corresponding tasks by the Delphi expert panel, time studies were planned for each participating hospital to quantify time spent in these areas to inform development of the workload productivity model. For the purposes of this study, time studies of the identified operational process categories and corresponding tasks were focused by preparation type. Preparation types were selected as a tool for measurement because dispensation time for a preparation comprises the combined times of operational tasks within an operational process. These preparation types were as follows: nonhazardous unit dose, hazardous unit dose, nonhazardous oral syringe, nonhazardous intravenous (IV) premixture, hazardous oral syringe, nonhazardous IV syringe, nonhazardous IV admixture, hazardous IV medications, and total parenteral nutrition (TPN). Data on time spent preparing as well as time spent verifying each preparation type were collected. For the purposes of this study, “product verification” was defined as the amount of time it takes a pharmacist to operationally check the product—from the time that they pick up the product to the time they sign off on the product, whether done electronically or manually. A point person from each site was selected to disseminate information from the research team to the operational pharmacists and technicians who internally collected observational time studies data on these processes. Seventeen operational areas across 8 hospital sites and infusion clinics participated in time studies conducted over 6 weeks from September 21 to October 31, 2020.

Sites completed both group and individual trainings to ensure the data collection process was understood and accurate data were collected. All sites collected data manually through time stamping, and these data were either recorded directly into an electronic spreadsheet or were transcribed to a spreadsheet after being written on a template that was provided by the study team. Additionally, 3 pharmacies at UNC Medical Center (UNCMC) (ie, the pediatric satellite pharmacy, central inpatient pharmacy, and cancer hospital infusion pharmacy) also collected data through their automation technologies (ie, barcode-assisted dispense preparation/verification, carousels, and gravimetric compounding system). To prevent the study results from being biased due to overrepresentation of a single site in the collected data, the amount of automated data utilized within the model for a single site did not exceed the total percentage of compounds created at that site in relation to the total percentage of compounds created throughout the system. For example, the UNCMC pediatric satellite pharmacy compounds 69% of nonhazardous IV syringes within the system; therefore, the amount of automated data for nonhazardous IV syringes prepared by the pediatric satellite (n = 43) that was used included in the model was 69% of the total amount of collected time study data for nonhazardous IV syringe doses (n = 62). Furthermore, automated data collected and utilized within the model were chosen at random from the total automated data available in order to decrease the risk of selection bias.

Data were collected weekly from each operational area. Median and interquartile range were calculated for the time spent on tasks within process groupings. To conserve limited resources at smaller institutions, a Kruskal-Wallis test (1-way analysis of variance on ranks) was performed each week to determine if the time spent on each task was significantly different between institutions. If different, pairwise Wilcoxon rank-sum tests with Bonferroni corrections were leveraged to identify outlier institutions that required individual follow-up.

Sites were instructed to discontinue time studies data collection on some dispense types prior to the end of 6 weeks because there was not a significant difference between the median and interquartile range times. At the end of week 5, the researchers allowed sites to stop time study data collection for preparation/compounding and verification of nonhazardous IV piggyback (IVPB) admixtures, premixed IVPB preparations, and hazardous unit dose products. Time study data collection was also stopped for preparation of nonhazardous unit dose products.

Based on the time studies, a weight was generated for each dispense type. This was done by calculating the median time per dispense necessary to complete medication preparation and product verification and then comparing and weighting that time in relation to all other dispense types. The process that took the least amount of time had a weight set at 1.0, and all other weights calculated were in comparison to that time. These weights were presented to the original stakeholder group of content experts to validate the data.

The developed model was then applied to 1 year of historic dispense data from December 29, 2019, through December 26, 2020. This application was to visualize the unique mix of medications and workload at each institution. A ratio of weighted work divided by the total number of dispenses (weighted operational ratio comparison [WORC] score) was also calculated to compare weighted workload between sites.

Results

After completion of time studies, the median time spent on preparation and median time spent on product verification were totaled to calculate a median total time per dispense for each preparation time (Table 1). Nonhazardous unit dose preparations required the least total time (71 seconds) per dispense, and TPNs required the most total time (1,078 seconds). A weighted workload calculation was then created by a time-based weight that was generated by calculating the median time taken for the entire workload process per dispense. In other words, weighting was determined by the ratio of total time for a preparation type to the total time for the preparation type requiring the least total time (ie, 71 seconds for nonhazardous unit dose) per dispense. The expert panel of operational leaders reviewed and validated the median total time by preparation type and corresponding model weighting score.

Table 1.

Weighted Workload Model

Preparation Type	Preparation Time, Median (IQR), seconds	Verification Time, Median (IQR), seconds	Total Time, Median (IQR), seconds	Model Weighting
Nonhazardous unit dose	60 (29)a	11 (9)a	71 (38)a	1.0
Hazardous unit dose	34 (33)a	21 (9)a	55 (42)a	1.11
Nonhazardous oral syringe	53	29	82	1.15
Nonhazardous IV premix	39	43	82	1.15
Hazardous oral syringe	95	10	105	1.47
Nonhazardous IV syringe	192	33	225	3.16
Nonhazardous IV admixture	386	53	439	6.18
Hazardous IV medications	473	37	494	7.18
TPN	980	98	1,078	15.18

Abbreviations: IQR, interquartile range; IV, intravenous; TPN, total parenteral nutrition.

^aTime logged via automation. Due to large discrepancies observed between time study data and automated data, automated times were used to calculate model weighting for preparation type.

Specific to both nonhazardous and hazardous unit dose times, large discrepancies between times were observed when comparing system data to automated data collected at UNCMC. A system work group consisting of operational experts and site point people reviewed both the system data and automated time data (automated time medians are shown in parenthesis next to the system median times in Table 1). This group validated the automated times for nonhazardous unit dose medications and hazardous unit dose medications to create the model weighting. These automated median total times were then used to calculate the model weighting for nonhazardous unit dose and hazardous unit dose preparations.

Model weight scores were used to calculate WORC scores by dividing weighted work by total number of dispenses. Table 2 summarizes the WORC scores, allowing for comparison of workload among sites. An example application of the model is seen in Figure 1 and Figure 2. Figure 1 shows that Nash’s Central Inpatient Pharmacy had a lower WORC score than Caldwell’s Central Inpatient Pharmacy. The model can be applied to the data to determine the reason for these differences. While Figure 1 demonstrates a WORC score showing that the sum of dispenses and weighted work are related, Figure 2 demonstrates that major differences in weighted workload are seen with nonhazardous IV medications. Figure 1 and Figure 2 show only 2 selected examples of thousands of combinations within single operational areas and across any and all operational areas to explore and compare weighted workload across the health system by dispense code, weighted work, and ratio of weighted work to total sum of dispenses at each site.

Table 2.

Weighted Workload Model Application

Operational Area	Ratio of Weighted Work to Sum of Dispenses Over 52-Week Period (WORC Score)
Caldwell UNC Health Central Inpatient Pharmacy	2.88
Chatham Hospital Central Inpatient Pharmacy	2.16
Johnston-Clayton Central Inpatient Pharmacy	1.90
Johnston-Smithfield Central Inpatient Pharmacy	2.17
Nash UNC Health Central Inpatient Pharmacy	1.15
Nash UNC Health Outpatient Infusion Center	6.37
Pardee UNC Health Central Inpatient Pharmacy	2.34
Pardee UNC Health Cancer Center Infusion Pharmacy	5.04
UNC Rex Health Care Blue Ridge Infusion Center	6.20
UNC Rex Health Care Cary Infusion Center	6.32
UNC Rex Health Care Central Inpatient Pharmacy	1.91
UNC Rex Health Care Garner Infusion Center	6.62
UNC Rex Health Care Oncology Infusion Center	5.99
UNC Rex Health Care Wakefield Infusion Center	6.51
UNC Medical Center	2.37
UNC Hillsborough Central Inpatient Pharmacy	2.07
UNCMC McCreary Infusion Center	6.43
UNCMC Children’s Raleigh Infusion Center	5.99
UNCMC Pittsboro Infusion Center	5.92
UNCMC Rockingham Infusion Center	5.71
UNCMC Therapeutic Infusion Center	6.06
Wayne UNC Health Central Inpatient Pharmacy	1.92

Abbreviations: UNC, University of North Carolina; UNCMC, University of North Carolina Medical Center; WORC, weighted operational ratio comparison.

Figure 1.

Ratio comparison of sum of weighted work to number of dispenses between Caldwell Inpatient Pharmacy and Nash Inpatient Pharmacy.

Figure 2.

Comparison of nonhazardous IV weighted workload and sum of number of dispenses between Caldwell Inpatient Pharmacy and Nash Inpatient Pharmacy.

Discussion

This study is the first to describe a process for developing an operational workload model that can be leveraged across different entities within a health system. By identifying operational workload activities occurring at diverse sites across the health system and conducting time studies, a health-system weighted workload model was developed. The weighted workload model allows for the comparison of hospital and infusion center pharmacy operational areas in thousands of combinations within single and across any and all operational areas by dispense code, weighted work, and ratio of weighted work to total sum of dispenses at each site. While this research applied the model across a 52-week period, any scale and duration of time could be used to visualize changes over time (eg, over a fiscal year). This model alone does not suggest any actionable changes to be made within the health system but, rather, serves as a framework to evaluate trends. Further, this model allows for more accurate comparisons around workload to be made across sites. Overall, through visualization with the model and comparison of WORC scores, UNC Health will be able to identify potential areas for maximizing efficiencies, increasing productivity, and explaining differences in work and output. This process can be replicated at various health systems to account for diversity across their hospitals (eg, variation in size, automation, and acuity), visualize trends, and inform opportunities for optimization across the health system.

Through guidance from the research team, sites were able to perform standardized time studies to determine the amount of time it takes to prepare and verify the various dispense codes used within the healthcare system. The dedication of the operational areas ensured that the model was ultimately reflective of practice. Additionally, to the investigators’ knowledge it is the only model that attempts to have wide application across a health system. Previous models have primarily focused on single hospitals or even specific pharmacies within a hospital.^1-3 While the process for developing this weighted workload model is generalizable to other health systems, the results of this specific weighted workload model are unique to UNC Health and should not be applied unchanged to any other health system, as there may be variability in operational practices and roles.

Performing time studies allowed the research team to create a weighted workload model that is specific and applicable to the entire health system. Ultimately, utilizing this model will provide data to internally benchmark system workload, analyze trends, and determine causation for differences in weighted work or dispensation trends across operational areas. This data has the potential to inform areas for growth, increase operational efficiencies, and demonstrate value across operational areas.

While this study is one of the first to describe a process for developing a weighted workload health-system model that accounts for the diversity of various hospital sites, it is not without limitations. The initial implementation of time studies encountered significant barriers, including decreased staffing due to the coronavirus disease 2019 pandemic and thus low labor resources to perform time studies. Due to operational constraints surrounding staffing and reorganizing in the face of an international health crisis, time studies were minimized to 6 weeks rather than the originally planned 3 months. Additionally, a limited quantity of time studies occurred at many sites due to those same operational and staffing constraints, despite the model being created after patient volumes returned to a prepandemic state. Sites were asked to either collect a minimum of 200 time studies per week or collect data for a minimum of 4 hours per week. Further, sites were asked to collect time studies on different days each week if weekly collection occurred in one day. Allowing for this level of flexibility aimed to support sites in fulfilling the time study request as additional resources were available.

Additionally, time studies involved manual data collection, which is time consuming and prone to human error. It is possible there were differences in how time studies were collected across sites. It is possible there existed variation in the communication and training provided through the site point persons. However, the use of standardized training and training materials aimed to minimize this variability. Workloads can also vary from day to day; therefore, the model may have the potential for increased accuracy if time studies were performed for all dispenses or if a rotating schedule were used to increase variability of time study data collection.

The variability of time taken to complete tasks was, overall, more homogeneous for product verification compared to product preparation. This indicates preparation roles traditionally attributed to technicians have potential impact on overall model process weighting. This should be considered if the model is being utilized to address dispensation trends or to analyze trends in relation to staffing of pharmacists and technicians in operational areas across sites.

This study aimed to create a first step towards developing a comprehensive pharmacy operational workload model. There are still many items that are important to a final model that were beyond the scope of this project, including steps outside of product preparation and verification (ie, “end-to-end” workload [from purchasing to administration]). These items will be vital to compare workload to staffing/labor expenses. Of note, for the study it was assumed that all dose preparation tasks are “technician tasks” and all dose verification tasks are “pharmacist tasks”; this distinction would need to be addressed before it could be applied to labor expenses for states and institutions where “tech-check-tech” is being leveraged.

Finally, this study included sites with varying technology. While UNCMC was able to use validated data from electronic sources, no other site had the ability to collect time study data in this way. This limitation was addressed through review of the model with the stakeholder groups to ensure validity despite technological differences.

Conclusion

This research describes a novel way to weight operational workload across a diverse health system. It is the first model intended to comprehensively capture weighted workload tied to a dispense code across varying practice settings. This model can serve to better articulate the distinct differences between operational areas and can help UNC Health visualize changes that occur operationally over time. The model created will be utilized to evaluate and compare future trends.

Disclosures

This study was funded by a research grant from the ASHP Research and Education Foundation. This publication was supported by grant number UL1TR002489 from the National Center for Advancing Translational Sciences (NCATS) at the National Institutes of Health (NIH) and the Pharmacy Analytics and Outcomes Team at UNC Health. The authors have declared no potential conflicts of interest.

Previous affiliations

At the time of writing Dr. Petersen was affiliated with UNC Medical Center.